Computational Analogues of Constraints on Grammars: A Model of Syntactic Acquisition

A principal goal of modern linguistics is to account for the apparently rapid and uniform acquisition of syntactic knowledge, given the relatively impoverished input that evidently serves as the basis for the induction of that knowledge the so-called projection problem. At least since Chomsky, the usual response to the projection problem has been to characterize knowledge of language as a grammar, and then proceed by restricting so severely the class of grammars available for acquisition that the induction task is greatly simplified perhaps trivialized. consistent with our lcnowledge of what language is and of which stages the child passes through in learning it." [2, page 218] In particular, ahhough the final psycholinguistic evidence is not yet in, children do not appear to receive negative evidence as a basis for the induction of syntactic rules. That is, they do not receive direct reinforcement for what is no_..~t a syntactically well-formed sentence (see Brown and Hanlon [3] and Newport, Gleitman, and Gleitman [4] for discussion). Á If syntactic acquisition can proceed using just positive examples, then it would seem completely unnecessary to move to any enrichment of the input data that is as yet unsupported by psycholinguistic evidence. 2